Computational Fluid Dynamics Analysis Solves Pump Noise Problem
Pumping applications involving cooling water have been especially difficult to solve because of the presence of dissolved air inherent in a cooling tower sump. cc
Water that contains large amounts of dissolved air changes the apparent required net positive suction head (NPSH). In such applications, traditional correction techniques failed because the entire system was not analyzed and the source of the noise generation could not be pinpointed.
This article explains the steps taken to solve this type of problem for Dow Chemical's plant in Freeport, TX, through computational fluid dynamics (CFD) and recounts the results.
Noise problem observed
In 1991, four large 30 x 30-38-in double-suction cooling tower pumps operating at 36,000 gpm and 710 rpm were installed at the Dow plant. Although these pumps met performance specifications on the test stand, they proved to be noisy when installed. Sound power levels greater than 93 dbA were observed approximately 3 ft from the pump casing.
These pumps were operating under duress as indicated by noise as well as other signs. The impeller was removed after 12 months to 18 months of service and some cavitation damage was evident.
Noise is a chronic problem with many cooling water pump installations.
The relatively high vapor pressure of hot water and the presence of dissolved air are both factors that influence the onset and the degree of pump cavitation, which creates noise and damages impeller and casing surfaces.
Apart from the fluid, other items that should be examined when noise is observed are the system installation, which includes piping, sump, valves, elbows, foundation and piping supports; as well as the pump's selection, operating point, mechanical condition and design.
The suction chamber of these pumps wraps around a portion of the discharge volute. As the flow enters the suction chamber, it splits at the discharge volute and undergoes a series of turns as it approaches the impeller (Fig. 1). This is analogous to the flow through a series of elbows. Consequently, a non-uniform velocity/pressure distribution is imposed on the impeller inlet.
Leading to a cure
Because of the relative expense and difficulty of modifying the fluid or the system at the installation at Dow, the pump became the focus of the investigation.
To establish a good understanding of the fluid behavior within the pump, a CFD model was developed.
CFD programs are evolving into useful engineering tools that can predict fluid behavior within almost any geometry. Even if the fluid condition and system effects are not understood, a CFD model can provide a best-case scenario whereby the pump casing design and impeller design can be evaluated.
The process began by creating a 3-D computer-aided drafting (CAD) model of the suction inlet portion of the pump casing and impeller. The CAD model was then imported into the CFD package and a mesh was created within the fluid space. The discharge volute of the pump was not modeled because, according to measurements taken in the field, the source of noise was confined to the suction.
Three flowrates for water at nominal room temperature were examined, based on the typical operating range of the pump [i.e., best efficiency point (BEP), 50% BEP and 120% BEP].
The field installation was relatively simple. A short length of pipe and a diffuser connected the suction flange to a wall of an open sump. A positive head exists at the suction centerline. Because the entrance velocities were relatively low and there was nothing unusual in the suction piping, the boundary condition at the suction flange was specified as a uniform total pressure.
The casing (suction inlet), wearing ring and impeller were intended to be modeled together but, as a preliminary study, the impeller was modeled separately to isolate its influence on the pumped fluid.
Neither backflow nor separation of flow was identified within impeller passages. Recirculation would have been identified by backflow near the blade inlets if it were occurring. However, none was observed. Because the impeller was well behaved in the initial analysis, it was ruled out as a problem source.
The remainder of the analysis focused on the casing suction inlet with the volume between impeller hub and shroud included in the model to provide downstream effects. An angular momentum term was derived from the impeller analysis for the flow entering the impeller eye. Hence, this would be used as a downstream boundary condition in the inlet analysis.
The results of the casing analysis were interpreted graphically by creating plots on several key planes that pass through the model. These plots display static pressure, total pressure, velocity vectors within a plane and magnitude of velocity components perpendicular to a plane.
There are two major planes of interest. The first is perpendicular to the shaft and just outside of the impeller eye and reveals information about the flow as it enters the impeller. The second is parallel to the shaft and passes through its center, while lying at a 51 Degrees angle from the horizontal centerline of the impeller (Fig. 3). This plane displays the profile of the wearing ring, impeller and a nearly central portion of the suction volute and reveals information about how the flow approaches the impeller (Fig. 4).
In an ideal pump, a plot showing the magnitudes of velocity components entering straight into the impeller eye, just upstream of the impeller, should display one uniform flow field. However, this analysis indicates both radial and circumferential variations in inlet velocity (Fig. 6). Radial variations imply that flow entering between impeller blades will have a different speed near the hub than near the shroud. Circumferential variations imply that, at any given instant, one impeller blade will be loaded differently than the next.
There are some negative velocities in the lower region of the plot. Some of the flow is actually exiting the impeller in this plane, indicating that a certain amount of backflow is present. Similar plots were created for the 50% BEP and 120% BEP flow conditions. The flow variations were much less severe at the lower flow but increased at the higher flows.
There are also radial and circumferential variations of the static pressure in the same plane. One particular point of interest is the localized pressure zone found near the splitter. This location corresponds to a localized area of pitting (cavitation damage) found on the top half of the casing, thus providing some credibility to the accuracy of the analysis prior to any experimental verification.
Further upstream of the impeller, one side (right) of the suction passages' velocity distribution includes the flow traveling from the suction nozzle while the other (left) side of the distribution includes only flow that passes around the casing wearing ring and enters the impeller from the opposite side because of the location of the cut plane (Fig. 4).
The inner wall of the suction nozzle has a deep valley that lies just upstream of the wearing ring. This profile is created by the wall of the discharge volute that wraps around the impeller and passes through the suction passage. The valley is formed by the blending of the impeller housing and the discharge volute. Some of the flow entering the pump tends to follow the contour of this valley but then must "climb back out" on its way over the wearing ring. This redirection of flow causes some vortexing in the valley region.
For the portion of flow that follows this contour and continues over the wearing ring, there is yet another vortex within the impeller. The inside profile of the wearing ring creates a sharp discontinuity in the flow boundary. As a result, the fluid momentum carries the high-velocity fluid over the discontinuity but viscous forces cause a small portion of it to slow down and turn back into a somewhat stagnant region. Thus, the vortex is formed.
Thus far, the CFD analysis of the pump casing had revealed that the suction inlet design is less than optimal. The most significant conclusion at this stage of the project was that cavitation was likely to occur irrespective of the impeller design.
The casing wearing ring was chosen to be altered to smooth the flow because it is a replaceable part in this double-suction pump design.
The radius on the inside of the ring was increased significantly to avoid separation of flow as the fluid accelerates into the impeller eye. In addition, large tabs or ears were added on a portion of the ring circumference facing the suction nozzle. The desired effect was to create a bridge over the valley formed by the curvature of the discharge volute (compare Fig. 6 with Fig. 7).
Final analysis
The CFD model was updated to reflect the new wearing ring geometry. Then the analysis was rerun using the same boundary conditions as before.
The circumferential and radial variations in velocities normal to the plane of the impeller are much less prominent than they were with the original wearing ring. There is no longer a region of negative velocities, thus no flow exits the eye. Overall, the flow is much more uniform in velocity. As in the velocity plot, the variations in a revised static pressure distribution plot still exist but are much less severe.
Improvements can also be seen in the inlet velocity distribution (Compare Fig. 4 with Fig. 5, above). The extension on the new ring appears to serve its intended purpose. The entrance flow moves smoothly along the outer surface of the ring and accelerates much more smoothly into the impeller. The vortex patterns are no longer visible in the valley created by the discharge volute nor are they visible inside of the impeller. The magnitude of velocity is more nearly equal from one side of the impeller to the other as compared to the case with the original wearing ring.
The acceleration into the eye is less than with the original rings as a direct result of providing a large radius turn into the eye. These effects can also be seen in the static pressure plot (not shown), where the gradients are less severe than the originals. The isolated low-pressure region has collapsed into nothing more than a gradient near the inner surface of the wearing ring.
All of these analysis results point toward a better pump design for a quieter pump. Recirculation losses upstream of the impeller were virtually eliminated. Recirculation in the impeller eye has disappeared. The magnitude and direction of flow entering the impeller are more uniform in the circumferential and radial directions. The final CFD analysis has predicted a greatly improved inlet flow to the impeller and one would expect better pump performance.
A series of tests was conducted at the pump manufacturer's R&D facility using a spare pump identical to the other four in service at the Dow plant with a new bronze casing wearing ring to prove that the analysis was correct.
Efforts were also made to verify the analysis results by means of measurements within the suction flow field. A static pressure probe and pitot (total-static) probe were used to record pressures and velocities at specific locations within a plane that lay just outside of the wearing ring, based on the output of the CFD analysis.
The measured data were plotted along with the theoretical predictions and an excellent correlation was observed.
Modeling results
The primary cause of noise in the 30 x 30-38 horizontally split case double-suction pump was flow separation occurring in the suction chamber. Classic techniques would not have led to a solution.
CFD was used successfully to identify the problem area within the suction passage (volute suction nozzle), rule out impeller recirculation as a problem and predict an improvement with a new ring.
Testing validated the results of CFD analysis and the improvement obtained with a new wearing ring. There was an overall drop of 3 db, out of 96 db (dbA), in noise (Fig. 8) and a drop of peak-to-peak pulsation pressure ranges between 2 psi and 5 psi (Fig. 9) without affecting performance. Improved suction characteristics were obtained at low capacity.
CFD proved to be a useful tool in solving a difficult plant problem.
By John Pembroke, Senior Product Engineer, and Gene Sabini, Manager of Technology, Goulds Pumps, ITT Industries, Industrial Pump Div., Seneca Falls, NY; and David E. Littlefield, Senior Design Specialist, The Dow Chemical Co., Freeport, Texas.
Bibliography
Fraser, W.H., "Avoiding Recirculation in Centrifugal Pumps," Machine Design, 1982.
Fraser, W.H., Recirculation in Centrifugal Pumps, ASME, 1981.
Knapp, R.T., Daily J.W., and Hammitt F.G., Cavitation, McGraw-Hill, 1970.
McNulty, P.J. and Pearsall, I.S., "Cavitation Inception in Pumps," ASME Journal of Fluid Engineering, 1982.
Nelik, L. and Freeman, J., "Case 1: Cooling Water Pump Case Study-Cavitation Performance Improvement," Proceedings of the Thirteenth International Pump Users Symposium, Turbomachinery Laboratory, Texas A&M University, College Station, TX, 1996.
Stepanoff, A.J., Pumps and Blowers, John Wiley & Sons, 1965.
Sulzer, Sulzer Centrifugal Pump Handbook, Elsevier Science Publishers, 1989.
Vleming, D.J., A Method for Estimating the Net Positive Suction Required by Centrifugal Pumps, ASME, 1981.
System factors affect noise
Pumping system design and installation are often significant factors in a chronic noise situation.
Pump design features that affect noise/pressure pulsation include:
Cutwater clearance.
Cutwater clearance is influenced by the impeller-to-casing distance and by the shape and number of impeller vanes passing the casing tongue.
Shape of the suction collector passage.
If the flow makes an abrupt change in direction as it enters the eye of an impeller, separation occurs.
Non-uniform supply to the impeller.
Double-suction pump casings impose an asymmetrical fluid velocity distribution entering the impeller eye because of the nature of their design.
Cavitation at the wear rings.
Position and contour of double-suction inlet stop piece (splitter).
Rotor imbalance, which can cause a low frequency rumble.
As the pump vibrates, bearing and seals can wear, which leads to noise problems.
Installation features affecting noise/pressure pulsation include factors in the follow these areas:
1. Mechanical.
- Baseplate resonance;
- Poor foundation design; and
- Piping supports.
2. Piping/sump.
- Prerotation of the fluid entering the pump caused by poor intake or poor inlet piping design produces a low frequency rumble;
- Valves cavitating cause high frequency noise;
- Velocity of the fluid within the pipes can excite resonances and low frequency noise (pipe organ effects); and
- Horizontal elbow on double-suction pump suction causes non-uniform velocity/pressure distribution from side to side of double suction. In very extreme cases, cavitation damage has been observed on one side of the impeller and recirculation damage on the other. Similar problems occur when valves are incorrectly placed near the pump suction.
3. Driver/electrical noise.
- High-pitched noise caused by the electromagnetic fields excites alternating forces on the motor's rotor and stator;
- Variable frequency drives can also produce a high frequency noise caused by harmonics; and
- Noise emanating from fan-cooled motors.